Research

Reassessing enzyme kinetics: considering protease-as-substrate interactions in proteolytic networks

Proteases are enzymes that degrade proteins and play a major role in cellular homeostasis. Cysteine cathepsins are a family of lysosomal proteases involved in extracellular matrix remodeling. This family contains the most potent mammalian collagenase and elastase and are often found upregulated in tissue destructive diseases such as cancer, osteoporosis, atherosclerosis, rheumatoid arthritis, and tendinopathy. Since cathepsins are so potent and involved in many different pathologies, they have been a key target for pharmaceutical companies. However, although well-design, specific cathepsin inhibitors have been developed and proven effective at stopping disease progression, all have failed phase II and III clinical trials due to off-target side effects. We hypothesize that an incomplete understanding of how individual cathepsin species interact with each other could be causing problems appropriately dosing inhibitors for treatment. Previously our lab has shown that cathepsin S will preferentially degrade cathepsin K, in the presence of substrate, a phenomenon termed “cathepsin cannibalism”. My research has focused on using computational modeling to tease apart these cathepsin cannibalism interactions between cathepsin K, L, S, and V and verifying these interactions by designing and making cannibalism-resistant cathepsin mutants.

Mulit-cellular Engineered Living Systems Proteolysis

Locomoting biological machines are innovative cell-based soft robotic devices that have the capability to dynamically sense and respond to environmental stimuli. However with current designs after a couple weeks, the C2C12 cells have reduced contractility due to some unknown causes. My research has focused on characterizing the protease activity in these biological machines for cathepsin and matrix metalloproteinase (MMP) activities using gelatin zymography and Western blots, which we hypothesized as the cause for machine failure. I characterized the protease activity in these biological machines with and without electrically stimulated contraction and under different design parameters. Once characterized, the biological machine lifespan potential was modeling, so we can design strategies to control the lifespan of the machine.

Quantifying intra-tumor heterogeneity

Investigating the underlying evolutionary dynamics of clonal hematopoietic diseases such as myelodysplastic syndromes and chronic myelomonocytic leukemia is important to understand where the heterogeneous clinical presentation of these malignancies with a limited amount of genomic heterogeneity. To this end, my efforts have focused on quantifying heterogeneity in tumor samples, compared to healthy cases, in attempts to identify key phenotypic signatures that can ultimately be used prognostically in the clinical setting. Specifically, I have developed a pipeline for analyzing single cell RNA-sequencing data and establishing that a generalized diversity index can successfully and robustly discriminate between normal and malignant tissue. Furthermore, higher diversity indicates higher clonality and this trend is replicated in both solid and liquid tumor samples.

Computational framework of parameter estimation of mechanistic models in cancer

Much of mechanistic cancer modeling revolves around studying the long-term success of co-evolving, complex interaction-driven tumor growth combined with statistical outcomes of treatment. Often, the data available to parameter these mechanistic models is underpowered and it is difficult to accurate the patient-to-patient variability in small, cohort studies. To improve our confidence in parameterization of mechanistic models, I’ve developed a computational pipeline for parameterizing of both in vitro and clinical cohort ordinary differential equation (ODE)-based model systems. The first step is to build the ODE mathematical model, then use L1 Loss Regularization to shrink redundant parameters. Once redundant parameters have been removed, use step-wise parameterization of the ODE model to iteratively estimated parameters of different random subsamples of the full dataset. Selecting the median parameter set as the parameterized for the mechanistic model, in silico predictions can be simulated and compared preclinical data for validation.